Jump to

Skill · Keyword research

Keyword research.

Find the queries worth ranking for, then prioritize what to produce.

Find the queries worth ranking for, classify them by intent, cluster them into topics, and decide what to produce first. Four stages carry it: discover a wide net of candidates, classify each by search intent, cluster the ones that share a page, and prioritize by a composite score. Tool-agnostic; it works with any keyword tool or with Search Console alone.

Intent is the stage that decides everything downstream. The tool gives you the demand; the SERP gives you the intent, and matching the query to the right page type is what turns volume into outcomes.

Audience: SEO and content teams starting a new section, planning a content calendar, or hunting ranking opportunities on an existing site.

View the skill on GitHubBrowse the full catalog

The framework

Four stages from a wide net to a ranked list.

Each stage narrows the set. Get intent right in stage two or the rest is noise.

  1. 01Discover: cast a wide net from seed terms, competitor exports, Search Console queries, related searches, and customer and forum language. Aim for 200 to 500 candidates.
  2. 02Classify by intent: map every keyword to informational, navigational, commercial, or transactional. The tool gives the demand; the SERP gives the intent.
  3. 03Cluster: group keywords that should target one page, by SERP overlap (3 or more shared top-10 results) or topical relevance. One primary plus 5 to 15 secondaries per page.
  4. 04Prioritize: score each cluster on opportunity, difficulty, and strategic fit. Priority equals opportunity plus strategic fit minus difficulty; the top 20 percent get produced first.

Classify by intent

Four intents, each won by a page type.

Every keyword maps to one of four intents. When in doubt, look at what actually ranks: if page 1 is articles, the query is informational; if product pages, transactional.

  1. 01

    Informational

    Signaled by 'how to', 'what is', 'why'. Won by an article, guide, or tutorial.

  2. 02

    Navigational

    A brand or product name plus a modifier. Won by the brand homepage or a product page.

  3. 03

    Commercial

    Signaled by 'best', 'review', 'vs', 'comparison', 'alternatives'. Won by a listicle, comparison, or review.

  4. 04

    Transactional

    Signaled by 'buy', 'price', 'deal', 'near me', 'for sale'. Won by a product or category page.

Pairs with these platforms

Where the volume and difficulty estimates come from.

The skill is tool-agnostic and the required inputs name Ahrefs, Semrush, Moz, or Google Keyword Planner. Ahrefs and Semrush are the two keyword tools with integration pages here; either exports the volume, difficulty, and competitor keywords discovery needs.

Reference files

Two references that go alongside the SKILL.md.

  • references/keyword-research-template.md

    Spreadsheet column definitions and a markdown summary template for the top clusters.

  • references/intent-classification-guide.md

    Detailed examples of each of the four intent categories, including hybrid intents.

Browse all reference files on GitHub

Bridges to other skills

What the research feeds, and what it is not.

Research is the front of the pipeline. These cover the single-page work, the competitive frame, the audit of what exists, and the program the keywords feed.

  • Once the query is known

    seo-onpage

    Optimizing a single page whose target query is already known is on-page work. Research finds the query; on-page optimizes the page for it.

  • The competitive frame

    seo-competitor

    Comparing your site to named competitors across many dimensions is a wider exercise. Run research to find the keyword space, that skill to read the field.

  • What already exists

    seo-content-audit

    Research plans new content; auditing existing pages for keep, update, or prune decisions is the audit's job. The two meet when an audit surfaces a gap to fill.

  • The program above it

    content-strategy

    Plans the content program: pillars, calendar, governance. Keyword research feeds the pillars as a downstream filter rather than driving the strategy itself.

Open source under MIT

Read the SKILL.md on GitHub.

The skill source lives in the rampstackco/claude-skills repository alongside dozens of other skills covering the full lifecycle of brand and product work. This page is a structured overview; the SKILL.md is the source. MIT licensed.

View SKILL.mdBrowse the full catalog

Frequently asked questions.

How do I classify search intent?
Map every keyword to one of four intents: informational ('how to', 'what is'), navigational (a brand or product name plus a modifier), commercial ('best', 'review', 'vs', 'alternatives'), or transactional ('buy', 'price', 'near me'). The keyword tool's volume tells you the demand, but the SERP tells you the intent. When in doubt, look at what is actually ranking: if page 1 is articles, the query is informational; if it is product pages, the query is transactional. Treat hybrid intents (like 'best running shoes') as their dominant intent and note the modifier.
How many keywords should discovery surface?
Around 200 to 500 candidates for a typical content sprint, and more if you are planning a year of content. Cast a wide net across seed terms from the brief, competitor keyword exports, Search Console queries (the page-1 and page-2 queries an existing site already ranks for), related searches and 'People also ask' in real SERPs, and the language customers and forums actually use. Then deduplicate and clean before classifying.
How do I cluster keywords?
Group the keywords that should target the same page. Two clustering approaches work together: SERP overlap (if two keywords share at least 3 of the top 10 results, they target the same page, which is mechanical and reliable) and topical relevance (grouping by the underlying topic rather than word overlap). A typical cluster has one primary keyword (highest volume, broadest intent) and 5 to 15 secondaries on a single page. Treating clusters as one-keyword-per-page leads to thin, cannibalized content.
How do I prioritize what to produce?
Score each cluster on three dimensions, 1 to 5: opportunity (volume, click potential, conversion potential), difficulty (the authority, backlinks, and content depth of the top results, plus whether SERP features compete with organic), and strategic fit (does it serve the audience and support the positioning). The priority score is opportunity plus strategic fit minus difficulty. Rank the clusters and produce the top 20 percent first. Do not chase volume without intent: a 10,000-volume informational keyword does not drive purchases.
Do I need a paid keyword tool?
It helps with volume and difficulty estimates, but the skill still works using SERP inspection and Search Console data alone, with rougher volume numbers. For an existing site, Search Console is especially valuable: it surfaces the queries the site already ranks for on pages 1 and 2, which are the easiest wins and are frequently queries the site never deliberately targeted.